Posted by Chris Lundie from dialin23.eagle.ca (206.186.242.232)on March 13, 1998 at 01:32:42:
RRBench was compiled with LCC. I heard that it wasn't very efficient, so I tried compiling RRBench with DJGPP. (and the RSXNTDJ add-on) DJGPP is known for its speed.
Here are the results from both compilers. As you can see, the fill-rate tests are almost identical, but the CPU-intense tasks can be up to twice as fast with DJGPP. I guess now I know which compiler to choose for my next project :)
My machine is Pentium 166, Stealth 2 (50/100).
DJGPP version. (no sync)
StarBlobs: 24.10 frames/sec
Fillrate: 49.80 frames/sec
Triangle low1: 171.65 frames/sec
Triangle low2: 163.70 frames/sec
Triangle low3: 314.30 frames/sec
Triangle med1: 104.30 frames/sec
Triangle med2: 69.70 frames/sec
Triangle high1: 17.95 frames/sec
Triangle high2: 12.00 frames/sec
Original version. (no sync)
StarBlobs: 23.75 frames/sec
Fillrate: 48.90 frames/sec
Triangle low1: 95.05 frames/sec
Triangle low2: 83.95 frames/sec
Triangle low3: 154.15 frames/sec
Triangle med1: 58.35 frames/sec
Triangle med2: 42.60 frames/sec
Triangle high1: 10.45 frames/sec
Triangle high2: 7.60 frames/sec